基于HSV色彩空间与形状特征的交通灯自动识别  

Traffic Light Auto-recognition Based on HSV Color Space and Shape Feature

作　　者：朱永珍[1] 孟庆虎[1,2] 普杰信[1] 

机构地区：[1]河南科技大学信息工程学院,河南洛阳471003 [2]香港中文大学工程学院,香港99907

出　　处：《电视技术》2015年第5期150-154,共5页Video Engineering

基　　金：国家自然科学基金项目(61301230);河南省重点科技攻关计划项目(092102210293)

摘　　要：为了在复杂的环境中准确识别出交通灯信息,提出一种基于HSV色彩空间和形状特征的交通灯识别方法。该方法首先将图像的RGB色彩空间转换成HSV色彩空间,并根据HSV色彩空间中的H与V无关性,利用不同颜色的H阈值对图像进行分割提取候选区域;然后对原图像经预处理及灰度形态学操作后,利用Hough变换检测目标可能位置;最后把目标疑似位置与候选区域进行逻辑滤波融合,并对融合后图像依据颜色H值判定交通灯信息。该方法对480幅各种场景下的交通灯图片进行实验,结果表明该方法具有很强的鲁棒性、稳定性和高效性,能够较好地识别出交通灯。Traffic light is an important signal that ensures vehicle running smoothly and orderly, it also can improve safety of drivers who are color-blind. In order to recognize the traffic light correctly in complex environment, in this paper, traffic light recognition automatically will be described. Firstly, by converting the color space form RGB to HSV, the candidate regions of traffic light are extracted using the H ,which base on the independence between H and V in HSV color space. At the same time, RGB image is turned into gray and after preprocessing and gray-scale morphological operation for the gray image, by using a method on the basis of the Hough transform is applied to detect the region that approximate location of the circle shape. Finally,the suspected location of the target regions and candidate regions are fused by using logical filtering, color distribution combined with the position of circle shape to recognize the traffic light. In this process, contrast enhancement and morphological operations and so on are also used. This experiments show that this method is robustness, stability and efficiency.

关 键 词：交通灯识别 HSV色彩空间 H分割 HOUGH 图像融合 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]